Added ability to set scheduling policy and CPU affinity for silk multiplexer processes to protect them from resource starvation in case of many backends running in parallel.

Added support for ALTER COLUMN TYPE and ALTER COLUMN RENAME commands on distributed tables.

Fixed a bug in global horizon calculation that could lead to autovacuum freezes under load.

Fixed too strong locking in CREATE INDEX CONCURRENTLY command that could lead to a distributed deadlocks under load.

Added a whitelist approach to push down several safe stable functions: date_trunc, to_date, to_char (only immutable functions supported push down before).

Various other bug fixes.

shardmanctl shard: added replicas reinit command to forcefully reattach replicas to a shard.

shardmanctl set: new command to quickly set pgParameters.

shardmanctl forall --twophase: added automatic resolution for interrupted 2PC transaction.

shardmanctl (probackup) recover: added --schema-only restore flag and PITR mode.

Fixed freezes in rebalance process in case of lack of resources with reconnect and timeout logic.

Fixed freezes in switchover process in case of several sequential attempts to switch master forth and back.

Added support for arbitrary number of master and replica instances on one node in manual topology mode.

Added checkpoint on master nodes before pg_ctl stop to speedup cluster restart process.